Heart Disease Detection using Green Anaconda Optimization Selection and MLGConv-based LWTNet classification with Enhanced PeaFowl-Based Artificial Intelligence Algorithm

The leading cause of death worldwide is heart disease. To counter this, a Deep Learning (DL) system, a subset of Artificial Intelligence (AI), has been developed to lower mortality rates by utilizing clinical insights for early detection. This study presents a novel method for diagnosing heart disease in patients and improving their prognosis. The extensive analysis is supported by multiple datasets sourced from repositories like UCI and Kaggle and locations like Long Beach, Cleveland, Switzerland, and Hungary. Gaussian Noise Removal, Motion Noise Reduction, and Contact Loss Handling (CLH) are used with a Fuzzy Inference System (KL-based FIS) integrated with Kalman filtering during the data preprocessing stage. The state-of-the-art Green Anaconda Optimization (GAO) is used for feature selection, simulating the stages of exploration and exploitation seen in wild green anacondas (GAs). The Multi-Level Group Convolution Light Weight Transformer Network, or LWTNet, which is based on MLGConv, is the foundation of the heart disease classification model. MLGConv is a module that effectively maintains lower computational costs while simultaneously representing multi-level and multi-group features, thereby improving local feature diversity. The Light Former transformer block comes from MLGConv and uses the least processing power to capture global dependencies, resulting in the final LWTNet. The Enhanced PeaFowl Optimization Algorithm (EPFOA) is utilized in the hyperparameter tuning of the classifier model. Interestingly, the suggested model outperforms earlier versions, with a remarkable accuracy rate of 99.87% in a comparative study.